What's new in 0.44.0:

The key feature in this release:

 - upsmon gets smarter.

   upsmon can now poll more than one UPS all within the same process.
   This should be very good news for people with big servers, multiple
   power supplies, and other interesting configurations.

   *** Note: this means you no longer specify the UPS identifier and
   master or slave on the command line.  See the upsmon.conf for how
   to define the new "MONITOR" directives.

   upsmon now handles a UPS that disappears differently.  If it goes away
   while on battery, it's considered to be critical and no longer 
   contributes to the overall "power value" or number of supplies that are
   OK.  If this makes the power value drop below the minimum safe level,
   the shutdown process begins.

   Note that the UPS is given a short interval to return before being
   declared 'dead'.  See the included example upsmon.conf for more info.
   
   Also new is the concept of a forced shutdown.  The details are too
   lengthy for a release note, so here's the basic idea.  If the master
   has to shut down a UPS for a reason other than 'on battery, low
   battery', the slaves now have a way to know about it and shutdown
   cleanly.  See Changes and the docs subdirectory for the whole scoop.

Other new features:

 - The apcsmart driver now supports an alternate powerdown sequence
   to avoid a different race when the power returns.  This is mostly
   for newer models.  Enable it with -s.
   [Nigel Metheringham]

 - RH users: The RPM spec file now builds three packages:

   nut        - client software
   nut-server - server software
   nut-cgi    - CGI tools

   [Nigel Metheringham]

 - Most drivers now support the -d <num> feature to define how long to
   wait before powering off.  -d 0 exits without waiting.
   [Nigel Metheringham]

 - Using --with-modelpath during configure will install the model drivers
   to some directory other than the normal "bin".  [Nigel Metheringham]

 - Similarly, --with-cgipath will move just the CGI binaries somewhere
   else.  [Nigel Metheringham]

 - The Makefiles now recognize INSTALLROOT for easier package building.
   Use "INSTALLROOT=/foo/bar make install" and the entire tree will be
   installed relative to /foo/bar.  [Nigel Metheringham]

 - upslog now supports %ETIME% to print the epoch time.  [Chad Madison]

 - Best Power Patriot Pro models are now recognized by the bestups driver.
   [Jon Nelson]

 - The apcsmart driver now supports the 940-0095B driver.  Use 
   '-c 940-0095B' to enable the alternate cable mode.  [David Corbin]

New drivers:

 - dummyups: not a real driver, but useful for testing.  It provides a 
   simple GUI through Xdialog which lets you set status values.  It's 
   primarily used to test upsmon without stressing my poor UPSes any more.

 - bestfort: supports older Best Fortress (usually black) models.
   [Grant Taylor]

 - mustekups: supports Mustek Electronics models.  [Jeremy Maccelari]

 - genericups type 10: Victron Lite support.  [Daniel Prynych]
   Also see docs/cables/victron.txt for more information on these models.

Bug fixes:

 - Capability strings over 256 chars should no longer break the
   apcsmart driver.  [Nigel Metheringham]

 - RH6 boxes should again be able to link gd with the libX11 and libttf
   libraries which means upsimage will work again.  [Bill Carlson]

 - The programs no longer step on the NetBSD libc functions send and recv.
   These have been renamed to upssendchar and upsrecv, respectively.
   [Will Ferry]

 - The apcsmart driver handles drivers without capabilities better.
   [Will Ferry]

 - The fentonups driver should again report a manufacturer value.

 - The included clients (upsc, upsct, upsct2, upscmd) now handle UPSes
   that either don't exist, are stale according to upsd, or have no
   supported features without segfaulting or spewing odd error messages.

 - Calling apcsmart with just '-d 0' or any option without specifying
   a device no longer segfaults.  [David Corbin]

 - The Cyberpower Power99 type in the genericups driver should no longer
   report "LB OL" erroneously.  [Geoffrey Keating]

 - upsmon should clean up after itself when receiving SIGQUIT or SIGTERM.
   Previously it left sockets hanging around in TIME_WAIT.  [Boris Popov]

 - upsmon can now tell the difference between POLLFREQ and POLLFREQALERT.
   This simple parsing bug caused upsmon to occasionally go crazy and 
   poll upsd constantly.  This drove CPU usage disturbingly high as
   both processes were kept busy by all the activity.  [Boris Popov]

 - The models and clients should now install properly again, even on
   systems where the install script/program only handles one file at a
   time.  [Lars Balker Rasmussen]

What's next:

 - More upsmon improvements, specifically event handling.

   When things start happening, upsmon likes to tell you about it.  The
   focus will be on making all those actions configurable.  If you want to
   send walls, e-mails, cat an .au into /dev/audio, or whatever, it will
   be possible.  It will also be possible to squelch certain messages
   entirely.

Out there:

 - Lots of people want to do "early shutdowns" - namely, kicking off
   a certain response to the on battery event after a defined interval.
   This will be addressed.